Rooba
/
boinc
mirror of https://github.com/BOINC/boinc.git
1
0
Fork 0
Code Issues 1 Packages Projects Releases Wiki Activity

*** empty log message *** 

svn path=/trunk/boinc/; revision=10637
This commit is contained in:
David Anderson 2006-07-12 22:25:27 +00:00
parent d6cebe3380
commit defbae7e7d
3 changed files with 88 additions and 72 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

152
checkin_notes
View File

@ -7098,28 +7098,28 @@ David 30 June 2006
filesys.C
David 30 June 2006
- core client: ACTIVE_TASK::supports_graphics():
return false if process isn't executing.
(prevent GUI from showing "show graphics" button
for results with no process, or suspended process)
- core client: on Windows command-line, when you close the window,
quit_client() gets called,
but control never returns to boinc_main_loop(),
so gstate.quit_activities() (and other cleanup stuff) isn't called,
so apps don't exit.
If you then start the core client again in the next 30 sec,
and exit quickly, another copy of apps gets started,
which waits to acquire slot lock, then decides it's in
standalone mode and opens a graphics window! Very confusing.
- core client: ACTIVE_TASK::supports_graphics():
return false if process isn't executing.
(prevent GUI from showing "show graphics" button
for results with no process, or suspended process)
- core client: on Windows command-line, when you close the window,
quit_client() gets called,
but control never returns to boinc_main_loop(),
so gstate.quit_activities() (and other cleanup stuff) isn't called,
so apps don't exit.
If you then start the core client again in the next 30 sec,
and exit quickly, another copy of apps gets started,
which waits to acquire slot lock, then decides it's in
standalone mode and opens a graphics window! Very confusing.
I fixed this by moving all the cleanup code to a separate functions
(finalize()) and calling this directly from quit_client().
I fixed this by moving all the cleanup code to a separate functions
(finalize()) and calling this directly from quit_client().
By the way, why is main.C such a toxic waste dump?
By the way, why is main.C such a toxic waste dump?
client/
app_graphics.C
main.C
client/
app_graphics.C
main.C
David 30 June 2006
- message tweaks
@ -7363,11 +7363,11 @@ David 9 July 2006
cpu_sched.C
David 9 July 2006
- core client: improve "RPC succeeded" message
- core client: improve "RPC succeeded" message
client/
cs_scheduler.C
scheduler_op.C
client/
cs_scheduler.C
scheduler_op.C
David 9 July 2006
- core client: instead of calling compute_work_requests()
@ -7409,60 +7409,60 @@ Matt 10 July 2006
forum.inc
David 11 July 2006
- Feeder: an option so that apps have a "weight" (floating-pt)
that determines the proportions of work items inserted by the feeder.
I.e. if app 1 has weight .5 and app 2 has weight 1.5,
then we'll send 3 times as many results for app 2 as app 1
- Feeder/scheduler: add an option "-reliable N":
This tags work items whose workunits are older than N days
as "need_reliable",
so that the scheduler can try to send them to "reliable" hosts
(those with expavg_credit > 70)
In no case send a "need_reliable" result to a host
with zero total credit.
- Feeder, with homogeneous redundancy:
when add a result with a sibling that's already sent
(or on a collision, if we learn that this is the case)
increment its infeasible_count.
This causes it to be sent quicker.
It's better to have the array full of results
that are not already committed to a platform.
- validator: add -sleep_interval cmdline arg
- Add TRANSITIONER_ITEM::res_received_time
(Nothing currently uses this - should we remove??)
- Feeder: an option so that apps have a "weight" (floating-pt)
that determines the proportions of work items inserted by the feeder.
I.e. if app 1 has weight .5 and app 2 has weight 1.5,
then we'll send 3 times as many results for app 2 as app 1
- Feeder/scheduler: add an option "-reliable N":
This tags work items whose workunits are older than N days
as "need_reliable",
so that the scheduler can try to send them to "reliable" hosts
(those with expavg_credit > 70)
In no case send a "need_reliable" result to a host
with zero total credit.
- Feeder, with homogeneous redundancy:
when add a result with a sibling that's already sent
(or on a collision, if we learn that this is the case)
increment its infeasible_count.
This causes it to be sent quicker.
It's better to have the array full of results
that are not already committed to a platform.
- validator: add -sleep_interval cmdline arg
- Add TRANSITIONER_ITEM::res_received_time
(Nothing currently uses this - should we remove??)
NOTE: this update requires a database change: see html/ops/db_update.php
NOTE: this update requires a database change: see html/ops/db_update.php
(from Kevin Reed)
(from Kevin Reed)
db/
boinc_db.C,h
schema.sql
html/ops/
db_update.php
sched/
feeder.C
sched_array.C
sched_send.C
sched_shmem.C,h
server_types.h
show_shmem.C
validator.C
db/
boinc_db.C,h
schema.sql
html/ops/
db_update.php
sched/
feeder.C
sched_array.C
sched_send.C
sched_shmem.C,h
server_types.h
show_shmem.C
validator.C
David 11 July 2006
- core client: fix bug where the first log flag in the
list was getting skipped over.
This was due to a design flaw in the new XML parsing code.
To work around this, I changed the format of
the cc_config.xml file so that options are
enclosed in <options> ... </options>
- core client: fix bug where the first log flag in the
list was getting skipped over.
This was due to a design flaw in the new XML parsing code.
To work around this, I changed the format of
the cc_config.xml file so that options are
enclosed in <options> ... </options>
client/
app.C
cs_statefile.C
log_flags.C,h
lib/
parse.C
client/
app.C
cs_statefile.C
log_flags.C,h
lib/
parse.C
Milos 11 July 2006
- Simple GUI Refresh
@ -7493,3 +7493,13 @@ Rom 12 July 2006
client/
main.C
David 12 July 2006
- core client: fix bug where sym links for project files
weren't getting written (missing fclose())
- core client: fix crash on project detach
(PROJECT object was being referenced after it was deleted)
client/
client_types.C
gui_rpc_server_ops.C

5
client/client_types.C
View File

@ -519,6 +519,10 @@ void PROJECT::write_project_files(MIOFILE& f) {
f.printf("</project_files>\n");
}
// write symlinks for project files.
// Note: it's conceivable that one physical file
// has several logical names, so try them all
//
int PROJECT::link_project_file(FILE_INFO* fip) {
char project_dir[256], path[256];
unsigned int i;
@ -531,6 +535,7 @@ int PROJECT::link_project_file(FILE_INFO* fip) {
FILE* f = boinc_fopen(path, "w");
if (!f) continue;
fprintf(f, "<soft_link>%s/%s</soft_link>\n", project_dir, fip->name);
fclose(f);
}
return 0;
}

3
client/gui_rpc_server_ops.C
View File

@ -211,7 +211,6 @@ static void handle_project_op(char* buf, MIOFILE& fout, const char* op) {
fout.printf("<error>must detach using account manager</error>");
return;
}
gstate.detach_project(p); // writes state file
// if project_init.xml refers to this project,
// delete the file, otherwise we'll just
@ -225,6 +224,8 @@ static void handle_project_op(char* buf, MIOFILE& fout, const char* op) {
);
}
}
gstate.detach_project(p); // writes state file; deletes p
gstate.request_schedule_cpus("project detached by user");
} else if (!strcmp(op, "update")) {
p->sched_rpc_pending = true;